This week we reviewed 210 articles from 21 sources. Here is what mattered.
The White House reversed course on Anthropic export controls, Meta's red-teaming contractors got caught posing as teens, and Mistral shipped a code agent that saturates PutnamBench. HP inked an OpenAI partnership, Ford admitted AI fell short and rehired veteran engineers, and SpaceX showed investors hardware.
Official Highlights
Anthropic redeployed Claude Fable 5 after the White House lifted export controls on Mythos and Fable models, adding a cybersecurity classifier that blocks a known jailbreak technique. Read more
HP scales its Frontier partnership with OpenAI to deploy AI across customer experiences, software development, and enterprise operations. Read more
Mistral shipped Leanstral 1.5, an Apache-2.0 Lean 4 code agent that solves 587 of 672 PutnamBench problems. The 119B MoE activates 6.5B parameters per token. Read more
NVIDIA HORIZON evolves Git worktrees to solve RTL problems hands-free, hitting 100% completion across benchmarks by treating each problem as a versioned repository. Read more
NVIDIA's BioNeMo Agent Toolkit turns biomolecular models like OpenFold3 and DiffDock into callable skills with programmatically exposed failure modes, enabling automated fallback chains in drug-discovery pipelines. Read more
Anthropic launched Claude Science Beta, a multi-agent workbench that delegates genomics and proteomics tasks to domain specialists while a reviewer agent flags citation errors. Read more
From the Community
The DeepMind trio who built a poker AI now run EquiLibre Technologies, a Prague-based quant shop valued above $500 million. Read more
Meta contractors posed as teenagers to test rival chatbots on suicide, sex, and drugs, prompting questions about adversarial testing ethics at scale. Read more
Ford rehired veteran engineers after AI-driven design systems fell short of production quality, an admission from a manufacturing leader. Read more
Anthropic Claude Sonnet 5 narrows the agentic coding gap with Opus 4.8 at Sonnet-tier pricing; benchmarks and cost-per-task breakdowns are now live. Read more
LlamaIndex's legal-kb is the first open-source reference app to combine grep with versioned legal documents at scale, giving agents filesystem-style access via retrieve, find, read, and grep tools. Read more
Alibaba reportedly classified Claude Code as high-risk software and banned employee use. Read more
Alibaba's Page Agent runs as client-side JavaScript inside the webpage, reading the live DOM to execute natural-language commands without screenshots or backend rewrites. Read more
Structured PDF-to-JSON: a comparative guide to open-source extraction models in 2026, weighing accuracy and throughput for production pipelines. Read more
Featured This Week
SpaceX showed investors a handset-like AI device prototype before going public, according to TechCrunch. The hardware tests Musk's expansion into consumer wireless and pits Starlink's 4.5 million subscriber terminals against Apple's and Google's AI device distribution. The prototype reportedly integrates with existing Starlink infrastructure, giving it a network backbone that would take years to build from scratch. Whether it ships depends on FCC appetite for another Musk spectrum play. Read more
Send comments and story tips to tips@planet-ai.net. If you found this useful, share it with a colleague.
This Week in AI
Additional stories worth scanning. Title only — click through for the full piece.
New Google commercial imagines a Declaration of Independence written with help from AI — TechCrunch · 2026-07-04
The only AI glossary you’ll need this year — TechCrunch · 2026-07-03
Mark Zuckerberg tells staff that AI agents haven’t progressed as quickly as he’d hoped — TechCrunch · 2026-07-02
Goose, a New Gay Dating App, Appears to Be a Psyop — WIRED · 2026-07-01
Midjourney wants Hollywood studios to reveal the details of their AI usage — TechCrunch · 2026-07-04
The browser wars aren’t about search anymore — here are the best alternatives to Chrome and Safari — TechCrunch · 2026-07-03
Jersey Mike’s IPO illustrates how bad the AI hype has become — TechCrunch · 2026-07-02
OpenClaw is finally available on Android and iOS — TechCrunch · 2026-06-30
Gemini’s personalized AI image generation is now free for US users — TechCrunch · 2026-06-29
Google DeepMind Unionization Talks Are Off to a Rocky Start — WIRED · 2026-07-03
Meta quietly launches vibe-coded gaming app Pocket — TechCrunch · 2026-07-02
Ashton Kutcher leaving Sound Ventures to launch new VC firm with Morgan Beller — TechCrunch · 2026-07-01
Anthropic and Gov. Newsom forge deal allowing California government to use Claude at half price — TechCrunch · 2026-06-29
Anthropic is discussing a new custom chip with Samsung — TechCrunch · 2026-07-02
Boleslaw Wyslouch steps down as director of Laboratory for Nuclear Science — MIT News · 2026-07-01
Google introduces a faster, cheaper image generator with Nano Banana 2 Lite — TechCrunch · 2026-06-30
South Korean tech giants commit over $550B to ease ‘ RAMageddon’ — TechCrunch · 2026-06-29
You Can Now Sound the Alarm on AI Behaving Badly — WIRED · 2026-07-01
Nvidia competitor Etched hits $5B valuation, $1B in sales for AI chip — TechCrunch · 2026-06-30
Arena, the AI leaderboard everyone uses, is now a $100M business — TechCrunch · 2026-06-29
OpenAI proposed donating 5% of its equity to a US sovereign wealth fund — TechCrunch · 2026-07-02
Cloudflare’s new policy pushes AI companies to pay for publishers’ content — TechCrunch · 2026-07-01
Cursor now has a mobile app for guiding your coding agent on the go — TechCrunch · 2026-06-29
Microsoft launches its own AI deployment company with $2.5 billion commitment — TechCrunch · 2026-07-02
Anthropic Added a New Security Measure to Get Back Into the Trump Administration’s Good Graces — WIRED · 2026-07-01
This newsletter supports planet-ai.net, a curated aggregator for AI tutorials and official updates. Curated by Keith Larson.
No comments:
Post a Comment